2020-12-21 16:32:31

by Eddie James

Subject: [PATCH 1/2] hwmon: (pmbus) Add a NO_PEC flag to probe chips with faulty CAPABILITY

Some PMBus chips don't respond with valid data when reading the
CAPABILITY register. For instance the register may report that the
chip supports PEC when in reality it does not. For such chips, PEC
must not be enabled while probing the chip, so add a flag so that
device drivers can force PEC off.

Signed-off-by: Eddie James <[email protected]>
---
drivers/hwmon/pmbus/pmbus_core.c | 8 +++++---
include/linux/pmbus.h | 10 ++++++++++
2 files changed, 15 insertions(+), 3 deletions(-)

diff --git a/drivers/hwmon/pmbus/pmbus_core.c b/drivers/hwmon/pmbus/pmbus_core.c
index 192442b3b7a2..3de1657dde35 100644
--- a/drivers/hwmon/pmbus/pmbus_core.c
+++ b/drivers/hwmon/pmbus/pmbus_core.c
@@ -2204,9 +2204,11 @@ static int pmbus_init_common(struct i2c_client *client, struct pmbus_data *data,
}

/* Enable PEC if the controller supports it */
- ret = i2c_smbus_read_byte_data(client, PMBUS_CAPABILITY);
- if (ret >= 0 && (ret & PB_CAPABILITY_ERROR_CHECK))
- client->flags |= I2C_CLIENT_PEC;
+ if (!(data->flags & PMBUS_NO_PEC)) {
+ ret = i2c_smbus_read_byte_data(client, PMBUS_CAPABILITY);
+ if (ret >= 0 && (ret & PB_CAPABILITY_ERROR_CHECK))
+ client->flags |= I2C_CLIENT_PEC;
+ }

/*
* Check if the chip is write protected. If it is, we can not clear
diff --git a/include/linux/pmbus.h b/include/linux/pmbus.h
index 1ea5bae708a1..9bdc8a581b03 100644
--- a/include/linux/pmbus.h
+++ b/include/linux/pmbus.h
@@ -34,6 +34,16 @@
*/
#define PMBUS_WRITE_PROTECTED BIT(1)

+/*
+ * PMBUS_NO_PEC
+ *
+ * Some PMBus chips don't respond with valid data when reading the CAPABILITY
+ * register. In this case, the register may report that the chip supports PEC
+ * with bit 7 (PB_CAPABILITY_ERROR_CHECK) when in reality it's not supported.
+ * For such chips, PEC must not be enabled before probing the chip.
+ */
+#define PMBUS_NO_PEC BIT(2)
+
struct pmbus_platform_data {
u32 flags; /* Device specific flags */

--
2.27.0

2020-12-21 18:37:35

by Eddie James

Subject: Re: [PATCH 1/2] hwmon: (pmbus) Add a NO_PEC flag to probe chips with faulty CAPABILITY

On Mon, 2020-12-21 at 08:54 -0800, Guenter Roeck wrote:
> On 12/21/20 8:30 AM, Eddie James wrote:
> > Some PMBus chips don't respond with valid data when reading the
> > CAPABILITY register. For instance the register may report that the
> > chip supports PEC when in reality it does not. For such chips, PEC
> > must not be enabled while probing the chip, so add a flag so that
> > device drivers can force PEC off.
> >
>
> I think the flag should indicate that the capability register
> shall not be read/used. That the capability register is currently
> only used to check for PEC is secondary. We might,for example,
> start using it to check for alert support or to check the numeric
> format.

OK, that makes sense. I'll rename the flag in v2, how does
PMBUS_NO_CAPABILITY sound?

Thanks for the quick reply,
Eddie

>
> Thanks,
> Guenter
>
> > Signed-off-by: Eddie James <[email protected]>
> > ---
> > drivers/hwmon/pmbus/pmbus_core.c | 8 +++++---
> > include/linux/pmbus.h | 10 ++++++++++
> > 2 files changed, 15 insertions(+), 3 deletions(-)
> >
> > diff --git a/drivers/hwmon/pmbus/pmbus_core.c
> > b/drivers/hwmon/pmbus/pmbus_core.c
> > index 192442b3b7a2..3de1657dde35 100644
> > --- a/drivers/hwmon/pmbus/pmbus_core.c
> > +++ b/drivers/hwmon/pmbus/pmbus_core.c
> > @@ -2204,9 +2204,11 @@ static int pmbus_init_common(struct
> > i2c_client *client, struct pmbus_data *data,
> > }
> >
> > /* Enable PEC if the controller supports it */
> > - ret = i2c_smbus_read_byte_data(client, PMBUS_CAPABILITY);
> > - if (ret >= 0 && (ret & PB_CAPABILITY_ERROR_CHECK))
> > - client->flags |= I2C_CLIENT_PEC;
> > + if (!(data->flags & PMBUS_NO_PEC)) {
> > + ret = i2c_smbus_read_byte_data(client,
> > PMBUS_CAPABILITY);
> > + if (ret >= 0 && (ret & PB_CAPABILITY_ERROR_CHECK))
> > + client->flags |= I2C_CLIENT_PEC;
> > + }
> >
> > /*
> > * Check if the chip is write protected. If it is, we can not
> > clear
> > diff --git a/include/linux/pmbus.h b/include/linux/pmbus.h
> > index 1ea5bae708a1..9bdc8a581b03 100644
> > --- a/include/linux/pmbus.h
> > +++ b/include/linux/pmbus.h
> > @@ -34,6 +34,16 @@
> > */
> > #define PMBUS_WRITE_PROTECTED BIT(1)
> >
> > +/*
> > + * PMBUS_NO_PEC
> > + *
> > + * Some PMBus chips don't respond with valid data when reading the
> > CAPABILITY
> > + * register. In this case, the register may report that the chip
> > supports PEC
> > + * with bit 7 (PB_CAPABILITY_ERROR_CHECK) when in reality it's not
> > supported.
> > + * For such chips, PEC must not be enabled before probing the
> > chip.
> > + */
> > +#define PMBUS_NO_PEC BIT(2)
> > +
> > struct pmbus_platform_data {
> > u32 flags; /* Device specific flags */
> >
> >

2020-12-21 18:47:23

by Guenter Roeck

Subject: Re: [PATCH 1/2] hwmon: (pmbus) Add a NO_PEC flag to probe chips with faulty CAPABILITY

On 12/21/20 8:30 AM, Eddie James wrote:
> Some PMBus chips don't respond with valid data when reading the
> CAPABILITY register. For instance the register may report that the
> chip supports PEC when in reality it does not. For such chips, PEC
> must not be enabled while probing the chip, so add a flag so that
> device drivers can force PEC off.
>

I think the flag should indicate that the capability register
shall not be read/used. That the capability register is currently
only used to check for PEC is secondary. We might,for example,
start using it to check for alert support or to check the numeric
format.

Thanks,
Guenter

> Signed-off-by: Eddie James <[email protected]>
> ---
> drivers/hwmon/pmbus/pmbus_core.c | 8 +++++---
> include/linux/pmbus.h | 10 ++++++++++
> 2 files changed, 15 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/hwmon/pmbus/pmbus_core.c b/drivers/hwmon/pmbus/pmbus_core.c
> index 192442b3b7a2..3de1657dde35 100644
> --- a/drivers/hwmon/pmbus/pmbus_core.c
> +++ b/drivers/hwmon/pmbus/pmbus_core.c
> @@ -2204,9 +2204,11 @@ static int pmbus_init_common(struct i2c_client *client, struct pmbus_data *data,
> }
>
> /* Enable PEC if the controller supports it */
> - ret = i2c_smbus_read_byte_data(client, PMBUS_CAPABILITY);
> - if (ret >= 0 && (ret & PB_CAPABILITY_ERROR_CHECK))
> - client->flags |= I2C_CLIENT_PEC;
> + if (!(data->flags & PMBUS_NO_PEC)) {
> + ret = i2c_smbus_read_byte_data(client, PMBUS_CAPABILITY);
> + if (ret >= 0 && (ret & PB_CAPABILITY_ERROR_CHECK))
> + client->flags |= I2C_CLIENT_PEC;
> + }
>
> /*
> * Check if the chip is write protected. If it is, we can not clear
> diff --git a/include/linux/pmbus.h b/include/linux/pmbus.h
> index 1ea5bae708a1..9bdc8a581b03 100644
> --- a/include/linux/pmbus.h
> +++ b/include/linux/pmbus.h
> @@ -34,6 +34,16 @@
> */
> #define PMBUS_WRITE_PROTECTED BIT(1)
>
> +/*
> + * PMBUS_NO_PEC
> + *
> + * Some PMBus chips don't respond with valid data when reading the CAPABILITY
> + * register. In this case, the register may report that the chip supports PEC
> + * with bit 7 (PB_CAPABILITY_ERROR_CHECK) when in reality it's not supported.
> + * For such chips, PEC must not be enabled before probing the chip.
> + */
> +#define PMBUS_NO_PEC BIT(2)
> +
> struct pmbus_platform_data {
> u32 flags; /* Device specific flags */
>
>

2020-12-21 18:55:36

by Guenter Roeck

Subject: Re: [PATCH 1/2] hwmon: (pmbus) Add a NO_PEC flag to probe chips with faulty CAPABILITY

On 12/21/20 10:32 AM, Eddie James wrote:
> On Mon, 2020-12-21 at 08:54 -0800, Guenter Roeck wrote:
>> On 12/21/20 8:30 AM, Eddie James wrote:
>>> Some PMBus chips don't respond with valid data when reading the
>>> CAPABILITY register. For instance the register may report that the
>>> chip supports PEC when in reality it does not. For such chips, PEC
>>> must not be enabled while probing the chip, so add a flag so that
>>> device drivers can force PEC off.
>>>
>>
>> I think the flag should indicate that the capability register
>> shall not be read/used. That the capability register is currently
>> only used to check for PEC is secondary. We might,for example,
>> start using it to check for alert support or to check the numeric
>> format.
>
> OK, that makes sense. I'll rename the flag in v2, how does
> PMBUS_NO_CAPABILITY sound?
>

sgtm

Thanks,
Guenter

> Thanks for the quick reply,
> Eddie
>
>>
>> Thanks,
>> Guenter
>>
>>> Signed-off-by: Eddie James <[email protected]>
>>> ---
>>> drivers/hwmon/pmbus/pmbus_core.c | 8 +++++---
>>> include/linux/pmbus.h | 10 ++++++++++
>>> 2 files changed, 15 insertions(+), 3 deletions(-)
>>>
>>> diff --git a/drivers/hwmon/pmbus/pmbus_core.c
>>> b/drivers/hwmon/pmbus/pmbus_core.c
>>> index 192442b3b7a2..3de1657dde35 100644
>>> --- a/drivers/hwmon/pmbus/pmbus_core.c
>>> +++ b/drivers/hwmon/pmbus/pmbus_core.c
>>> @@ -2204,9 +2204,11 @@ static int pmbus_init_common(struct
>>> i2c_client *client, struct pmbus_data *data,
>>> }
>>>
>>> /* Enable PEC if the controller supports it */
>>> - ret = i2c_smbus_read_byte_data(client, PMBUS_CAPABILITY);
>>> - if (ret >= 0 && (ret & PB_CAPABILITY_ERROR_CHECK))
>>> - client->flags |= I2C_CLIENT_PEC;
>>> + if (!(data->flags & PMBUS_NO_PEC)) {
>>> + ret = i2c_smbus_read_byte_data(client,
>>> PMBUS_CAPABILITY);
>>> + if (ret >= 0 && (ret & PB_CAPABILITY_ERROR_CHECK))
>>> + client->flags |= I2C_CLIENT_PEC;
>>> + }
>>>
>>> /*
>>> * Check if the chip is write protected. If it is, we can not
>>> clear
>>> diff --git a/include/linux/pmbus.h b/include/linux/pmbus.h
>>> index 1ea5bae708a1..9bdc8a581b03 100644
>>> --- a/include/linux/pmbus.h
>>> +++ b/include/linux/pmbus.h
>>> @@ -34,6 +34,16 @@
>>> */
>>> #define PMBUS_WRITE_PROTECTED BIT(1)
>>>
>>> +/*
>>> + * PMBUS_NO_PEC
>>> + *
>>> + * Some PMBus chips don't respond with valid data when reading the
>>> CAPABILITY
>>> + * register. In this case, the register may report that the chip
>>> supports PEC
>>> + * with bit 7 (PB_CAPABILITY_ERROR_CHECK) when in reality it's not
>>> supported.
>>> + * For such chips, PEC must not be enabled before probing the
>>> chip.
>>> + */
>>> +#define PMBUS_NO_PEC BIT(2)
>>> +
>>> struct pmbus_platform_data {
>>> u32 flags; /* Device specific flags */
>>>
>>>
>